Impersonate human decision making process: an interactive context-aware recommender system

Abstract

A considerable amount of information is quickly disseminated worldwide and users struggled to survive on such data tsunami. Context-recommender-aware systems (CAR) are then developed which enabling users to locate valuable and useful information from a large amount of disordered data. However, human decision-making contains multiple steps and a recursive loop, most users tend to adjust their decision many times instead of achieving the final decision-making immediately. Therefore, to replicate such a recursive process among multiple steps, the traditional CAR system should be altered as an interactive CAR (iCAR) system for improving the recommendation accuracy. In view of the deficiency in the present CAR, this study leads the concept of human-computer interaction in tradition CAR and establishes an interactive context-aware recommender System (iCAR). To validate the feasibility and applicability of the proposed iCAR system, a car rental website which is designed based on iCAR is shown as a demonstration. According to the car rental case shown, after couples of iterations, the decision criteria can be gradually clarified by the proposed algorithm of inferring engine. Also, iCAR can find users a car that most satisfies their requirements by using the contexts information. iCAR can improve the accuracy of traditional CAR system and provide user more precise recommendation results according to 3-dimensions information, including: user, item and context information. The iCAR system can be further expected to apply to various fields, such as online shopping or travel packages recommendations, to optimize recommendations results.